Python3中用于计算pow（a，b）的内置函数的复杂性

Question

使用迭代的函数以及n为偶数时ⁿ =（axa） ^{n / 2}和n为奇数时[[axa） ^{（n-1）/ 2} ] xa的事实是否会比python中的内置函数产生更好的结果3。

Answer 1

No, you will not be able to out perform Python built-in operations. The reason is fairly simple, those built-in operations run C code, in the case of CPython. You will not beat C code by writing Python code.

If we think about your case specifically, your solution would require to instantiate multiple integer or float objects, do comparions, multiply, all that in Python ! This is something the built-in power operator does not do, as Python has an instruction that computes the power in C and only instantiate one object, which is the value that you want.

We can see that by using dis to see python bytecode.

>>> import dis    
>>> dis.dis('a**b')
1         0 LOAD_NAME                0 (a)
          2 LOAD_NAME                1 (b)
          4 BINARY_POWER
          6 RETURN_VALUE

>>> dis.dis("""
r = 1
while b > 1:
    if b % 2:
        b = (b - 1) / 2
        a *= a
        r *= a
    else:
        b /= 2
        a *= a
r *= a""")

2         0 LOAD_CONST               0 (1)
          2 STORE_NAME               0 (r)

3         4 SETUP_LOOP              66 (to 72)
          6 LOAD_NAME                1 (b)
          8 LOAD_CONST               0 (1)
         10 COMPARE_OP               4 (>)
         12 POP_JUMP_IF_FALSE       70

4        14 LOAD_NAME                1 (b)
         16 LOAD_CONST               1 (2)
         18 BINARY_MODULO
         20 POP_JUMP_IF_FALSE       52

5        22 LOAD_NAME                1 (b)
         24 LOAD_CONST               0 (1)
         26 BINARY_SUBTRACT
         28 LOAD_CONST               1 (2)
         30 BINARY_TRUE_DIVIDE
         32 STORE_NAME               1 (b)

6        34 LOAD_NAME                2 (a)
         36 LOAD_NAME                2 (a)
         38 INPLACE_MULTIPLY
         40 STORE_NAME               2 (a)

7          42 LOAD_NAME                0 (r)
         44 LOAD_NAME                2 (a)
         46 INPLACE_MULTIPLY
         48 STORE_NAME               0 (r)
         50 JUMP_ABSOLUTE            6

9        52 LOAD_NAME                1 (b)
         54 LOAD_CONST               1 (2)
         56 INPLACE_TRUE_DIVIDE
         58 STORE_NAME               1 (b)

10       60 LOAD_NAME                2 (a)
         62 LOAD_NAME                2 (a)
         64 INPLACE_MULTIPLY
         66 STORE_NAME               2 (a)
         68 JUMP_ABSOLUTE            6
         70 POP_BLOCK

 11      72 LOAD_NAME                0 (r)
         74 LOAD_NAME                2 (a)
         76 INPLACE_MULTIPLY
         78 STORE_NAME               0 (r)
         80 LOAD_CONST               2 (None)
         82 RETURN_VALUE

I'll let you guess how disproportionately more efficient the built-in is.